Historian Jill Lepore argues that the current era of AI anxiety and platform politics has a long, ignored prehistory. Her central claim: between the 1950s and the 1990s, thinkers repeatedly warned that computing could replace democratic deliberation with what she calls the artificial state, government of people by machines, and nobody acted on those warnings.

The story starts in 1958, when advertising executive Edward L Greenfield and MIT political scientist Ithiel de Sola Pool proposed a People Machine: punch cards of election returns, census data and surveys feeding an information bank that would simulate the US electorate. In 1959 they founded the Simulmatics Corporation with IBM's Alex Bernstein, calling their method the A-bomb of the social sciences. The Democratic National Committee was the first client, and after John F Kennedy's 1960 victory the press credited Simulmatics as his secret weapon. The company went bankrupt in 1970, not because the model failed but because the data did not yet exist. The microtargeting it attempted in publishing and film only became feasible decades later at Amazon and Netflix.

Warnings came quickly. Eugene Burdick's 1964 novel The 480 depicted a thinly disguised Simulmatics meddling in a presidential election. Lewis Mumford's The Myth of the Machine (1966) warned that cybernetic intelligence would reduce humans to passive, machine-conditioned animals, and called technological determinism a radical misinterpretation of human development. Joseph Weizenbaum, who built the chatbot Eliza in 1966, wrote in 1976 that substituting computers for human functions involving respect, understanding and love was obscene, and refused to support work toward a fully automated battlefield.

Lepore also recounts the optimists. JCR Licklider estimated in 1965 that the sum of human knowledge was doubling yearly and would reach about a dozen petabytes by 2020; the real internet was measured in dozens of zettabytes. With Robert Taylor he predicted online interactive communities in 1968. Ithiel de Sola Pool predicted personalised news would atomise society and make rallying political support far harder. Chile's Project Cybersyn under Salvador Allende tried to run a socialist economy cybernetically from 1971 to 1973, until Pinochet's coup ended it. Stewart Brand promised in 1972 that the personal computer would bring power to the people.

The 1980s brought a Reagan campaign People Machine, Time's 1982 Machine of the Year, and Apple's 1984 Macintosh ad promising that 1984 would not be like 1984. Political theorist Langdon Winner countered that the computer revolution was conspicuously silent about its own ends and that the evidence pointed to more power and wealth for those who already had both. Meanwhile Usenet, opened in 1980, produced trolling and flame wars, and in 1983 white supremacist George Dietz launched the Liberty Bell Network, the first hate bulletin board, on an Apple IIe.

The 1990s, in Lepore's telling, settled the outcome. Silicon Valley absorbed Ayn Rand's radical individualism; Steve Jobs called Atlas Shrugged his guide in life. Newt Gingrich closed the Office of Technology Assessment in 1995. John Perry Barlow's 1996 Declaration of the Independence of Cyberspace told governments they had no sovereignty online, while alternative visions from Michael Hauben and Douglas Adams died with their authors in 2001. Clinton's 1996 Telecommunications Act delivered the largely unregulated internet the techno-libertarians wanted. Lepore's conclusion: the digital revolution played out this way because liberal democracy failed to limit corporate power over politics, not because the outcome was inevitable.